Study at a glance
Summary
Disclaimer: The following summary details independent academic research. Theia3D is intended for research, sports performance and educational use only. It is not cleared as a medical device and is not intended to be used for the diagnosis, treatment, mitigation or prevention of any disease, injury or medical condition.
Editor's note: Theia3D version 2023.2.0.4205, as stated in the paper.
Study Overview
Design
A concurrent comparison study using three simultaneously recorded camera systems, a marker-based reference system and two identically equipped markerless camera systems in different mounting positions, to separately assess markerless precision (agreement between the two markerless setups) and markerless accuracy (agreement of each markerless setup relative to the marker-based system) during jumping and cutting tasks, alongside a usability assessment of the markerless system software with novice and experienced motion-capture users.
Participants
Eleven competitive athletes (6 men, 5 women; mean age 22.2 years), intentionally recruited to represent a range of heights, body types, and skin tones, including skin tones that have been rarely reported in prior markerless validation research.
Data Collection
Participants completed a dynamic calibration trial followed by five repetitions each of four movement tasks: a maximum squat jump, a single-legged squat jump, a 45-degree side-step cut, and a drop vertical jump. All three camera systems, a marker-based system using 20 truss-mounted cameras and 41 reflective markers, a markerless system using 10 truss-mounted Qualisys Miqus cameras positioned at the same height as the marker-based system (approximately 3 meters), and an identical markerless system using 10 tripod-mounted cameras positioned lower (approximately 1.4 meters), recorded each movement simultaneously, with video data processed through Theia3D to produce kinematic output.
Analysis
Markerless precision was assessed by comparing joint angles from the two differently positioned markerless systems using intraclass correlation coefficients (ICC) and root mean square error (RMSE) at key movement timepoints and across full movement time-series curves. Markerless accuracy was assessed by comparing each markerless system against the marker-based reference system using Lin's concordance correlation coefficient, Pearson's correlation, and RMSE, both before and after applying a bias-adjusting shift to account for systematic offsets. Usability was assessed qualitatively with two novice and two experienced motion-capture users operating the markerless software.
Key Findings
- The two differently positioned markerless camera systems showed excellent precision with each other, with a mean intraclass correlation coefficient of 0.95 and mean RMSE of 1.91 degrees across all joint angles and tasks, and the 1.6-meter difference in camera height did not significantly affect these agreement metrics.
- Relative to the marker-based reference system, markerless accuracy was moderate, with a mean Lin's concordance correlation coefficient of 0.51 to 0.56 and mean RMSE of 8.48 to 9.26 degrees across joint angles and tasks, with the largest source of error concentrated in sagittal-plane hip flexion angle, particularly at higher degrees of flexion.
- After accounting for this systematic hip flexion bias with a correction shift, RMSE across the remaining joints and tasks dropped substantially, to 3.29 to 3.87 degrees, indicating that once this specific, identifiable offset was addressed, the two systems' remaining differences were considerably smaller.
- There was no significant difference in overall accuracy relative to the marker-based system between the higher, truss-mounted and lower, tripod-mounted markerless setups, based on concordance correlation and Pearson's correlation, though an initial, smaller difference in uncorrected RMSE between the two markerless setups was no longer significant after bias adjustment.
- Both novice users with no prior motion-capture experience and users experienced with other motion-capture systems rated the markerless software highly for ease of use and confidence, supporting its accessibility for personnel without specialized technical training.
What This Means for Movement Science Researchers and Sports Performance Assessment Teams
For researchers and performance staff working in sports settings, including organizations without dedicated motion-capture specialists, this study offers two specific, practically useful findings. First, markerless camera placement within a reasonable range, including positions lower than the system's own recommended camera height, did not meaningfully compromise the precision of the kinematic data produced in this study, a reassuring finding for facilities with physical space constraints that make ideal camera placement difficult to achieve. Second, the system's usability findings support its accessibility for staff without specialized motion-capture backgrounds, lowering a practical barrier to adoption for sports organizations. At the same time, the sagittal-plane hip flexion bias identified here reinforces a now-recurring pattern across markerless validation research: hip kinematics, and particularly hip flexion at larger angles, remain an area warranting careful interpretation, likely related to differences in how the hip joint center and pelvis segment are defined between markerless and marker-based models.
The Role of Theia3D in Movement Analysis
Theia3D was the markerless motion capture system directly evaluated in this study, both for its own internal consistency across different camera placements and for its accuracy relative to a marker-based reference system, during fast, explosive jumping and cutting tasks relevant to sports performance assessment. The study's inclusion of a dedicated camera-placement comparison, alongside the more typical markerless-versus-marker-based accuracy comparison, offers a distinct and practically useful contribution: direct evidence on how much flexibility exists in where cameras are positioned without compromising markerless data quality, a question most prior Theia3D validation research has not addressed directly.
Study Limitations and Future Directions
The authors note that differing definitions of the pelvis and hip joint center between the markerless and marker-based systems may affect comparisons between the two, and that marker-based estimates themselves are subject to known error from soft-tissue artifact affecting hip joint center estimation, which becomes increasingly unreliable at larger hip flexion angles, meaning the marker-based system used as the reference here is itself an imperfect benchmark for this specific joint and movement range. They note that because the underlying kinematic calculations within Theia3D are proprietary, it is difficult to identify the precise cause of the sagittal-plane hip angle differences observed. The authors also caution that because markerless software continues to be updated and improved, care should be taken when applying findings from earlier Theia3D validation research to current versions of the software. They identify further investigation into the markerless system's underlying modeling approach for the pelvis and hip, and continued evaluation across a broader and more diverse range of body types, heights, and skin tones, as useful next steps for this line of research.
Citation: Kwak, S. T., Riehm, C. D., Anand, M., Raphael, N., DiCesare, C. A., Schille, A., Hulburt, T. C., Zendler, J. M., Hybart, R. L., Buria, K. N., Trotter, N., & Myer, G. D. (2026). Assessing agreement and usability of kinematic measures between markerless and marker-based motion capture systems during jumping activities. Journal of Biomechanics, 194, 113065. https://doi.org/10.1016/j.jbiomech.2025.113065
Link: Read the full study
Does Camera Placement Affect the Accuracy of Markerless Motion Capture During Jumping?
This study of 11 competitive athletes finds that Theia3D showed excellent agreement between two differently positioned markerless camera setups and moderate accuracy relative to marker-based motion capture, with a notable bias concentrated in sagittal-plane hip flexion, and was rated easy to use by both novice and experienced motion-capture users. If your work involves sports performance assessment, talk to our team about applying Theia3D to your own facility.